LETTER: Ipsos’s poor track record must be kept in mind
Reporters should be aware that some polling organisations get things consistently wrong
10 November 2022 - 15:55
Support our award-winning journalism. The Premium package (digital only) is R30 for the first month and thereafter you pay R129 p/m now ad-free for all subscribers.
General election campaign posters are displayed on the road side in Pretoria in this April 18 2019 file photo. Picture: BLOOMBERG via GETTY IMAGES/WALDO SWIEGERS
Just three months ago, Ipsos published a poll that had the ANC at 42%, the DA at 11%, the EFF at 9%, others together at 12%, and various other responses such as “Won’t vote” at 26%. Now a new Ipsos poll, commissioned by Songezo Zibi’s Rivonia Circle, has the ANC at 41%, the DA at 18%, the EFF at 15%, others at 12% and “Don’t know” at 13% (“ANC support is at 41%, opinion poll shows”, November 9). If this poll is to be believed there have been wild swings in the opposition-supporting electorate in the last three months.
Unfortunately, media coverage of this poll has not mentioned Ipsos’s poor track record. In the run-up to the 2021 election its polls predicted doom and gloom for the DA, which the media merrily ran with. In August 2021, it had the DA at 17.9% and the ANC at 49%; in October 2021 it had the DA at 16.7% and the ANC at 30%. Neither of these poll results bore any resemblance to the actual election result.
While the latest Ipsos poll only samples registered voters, which is a huge improvement on its past methodology, there is no turnout modelling or weighting to account for differential turnout. Furthermore, it does not assign or exclude voters who responded “Don’t know”. Its exact demographic sample (age, geography, etc) is also not disclosed, so there is no way to determine the validity of the poll.
It is often stated that the only poll that matters is the one conducted on election day. Yet a stronger DA in the post-2021 election environment is evidently borne out by a series of by-elections where the DA has cleaned up in its traditional strongholds (in Durban, the DA recently won a by-election with 94% of the vote), clawing back support in areas where small parties briefly made inroads (including winning wards off the Patriotic Alliance) and growing in ANC strongholds.
The DA has won seven by-elections on the trot, which stands in stark contrast to the aftermath of the 2019 election environment.
When reporting on polling reporters should exercise a duty of care towards the polling results of organisations that have a track record of getting it consistently wrong. Good journalism requires that hard questions are asked about sampling, methodology and analysis, rather than mindlessly reporting polling numbers as fact.
Ashor Sarupen DA deputy campaign manager
JOIN THE DISCUSSION: Send us an email with your comments to letters@businesslive.co.za. Letters of more than 300 words will be edited for length. Anonymous correspondence will not be published. Writers should include a daytime telephone number.
Support our award-winning journalism. The Premium package (digital only) is R30 for the first month and thereafter you pay R129 p/m now ad-free for all subscribers.
LETTER: Ipsos’s poor track record must be kept in mind
Reporters should be aware that some polling organisations get things consistently wrong
Just three months ago, Ipsos published a poll that had the ANC at 42%, the DA at 11%, the EFF at 9%, others together at 12%, and various other responses such as “Won’t vote” at 26%. Now a new Ipsos poll, commissioned by Songezo Zibi’s Rivonia Circle, has the ANC at 41%, the DA at 18%, the EFF at 15%, others at 12% and “Don’t know” at 13% (“ANC support is at 41%, opinion poll shows”, November 9). If this poll is to be believed there have been wild swings in the opposition-supporting electorate in the last three months.
Unfortunately, media coverage of this poll has not mentioned Ipsos’s poor track record. In the run-up to the 2021 election its polls predicted doom and gloom for the DA, which the media merrily ran with. In August 2021, it had the DA at 17.9% and the ANC at 49%; in October 2021 it had the DA at 16.7% and the ANC at 30%. Neither of these poll results bore any resemblance to the actual election result.
While the latest Ipsos poll only samples registered voters, which is a huge improvement on its past methodology, there is no turnout modelling or weighting to account for differential turnout. Furthermore, it does not assign or exclude voters who responded “Don’t know”. Its exact demographic sample (age, geography, etc) is also not disclosed, so there is no way to determine the validity of the poll.
It is often stated that the only poll that matters is the one conducted on election day. Yet a stronger DA in the post-2021 election environment is evidently borne out by a series of by-elections where the DA has cleaned up in its traditional strongholds (in Durban, the DA recently won a by-election with 94% of the vote), clawing back support in areas where small parties briefly made inroads (including winning wards off the Patriotic Alliance) and growing in ANC strongholds.
The DA has won seven by-elections on the trot, which stands in stark contrast to the aftermath of the 2019 election environment.
When reporting on polling reporters should exercise a duty of care towards the polling results of organisations that have a track record of getting it consistently wrong. Good journalism requires that hard questions are asked about sampling, methodology and analysis, rather than mindlessly reporting polling numbers as fact.
Ashor Sarupen
DA deputy campaign manager
JOIN THE DISCUSSION: Send us an email with your comments to letters@businesslive.co.za. Letters of more than 300 words will be edited for length. Anonymous correspondence will not be published. Writers should include a daytime telephone number.
ANC electoral support set to decline further, poll shows
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.
Most Read
Published by Arena Holdings and distributed with the Financial Mail on the last Thursday of every month except December and January.